Targeted metagenomics and ecology of globally important uncultured eukaryotic

Marie L. Cuveliera,b, Andrew E. Allenc,1, Adam Moniera,1, John P. McCrowc, Monique Messiéa, Susannah G. Tringed, Tanja Woyked, Rory M. Welsha, Thomas Ishoeyc, Jae-Hyeok Leee, Brian J. Binderf, Chris L. DuPontc, Mikel Latasag, Cédric Guigandb, Kurt R. Bucka, Jason Hiltonb, Mathangi Thiagarajanc, Elisabet Calerc, Betsy Readh, Roger S. Laskenc, Francisco P. Chaveza, and Alexandra Z. Wordena,b,2

aMonterey Bay Aquarium Research Institute, Moss Landing, CA 95039; bRosenstiel School of Marine and Atmospheric Science, Miami, FL 33149; cJ. Craig Venter Institute, San Diego, CA 92121; dUS Department of Energy Joint Institute, Walnut Creek, CA 94598; eDepartment of Biology, Washington University, St. Louis, MO 63130; fDepartment of Marine Sciences, University of Georgia, Athens, GA 36072; gInstitut de Ciències del Mar (CSIC), E-08003 Barcelona, Spain; and hDepartment of Biological Sciences, California State University, San Marcos, CA 92096

Edited by David Karl, University of Hawaii, Honolulu, HI, and approved June 21, 2010 (received for review February 18, 2010)

Among , four major phytoplankton lineages are respon- Oceanic prymnesiophytes are thought to be small owing to high sible for marine ; prymnesiophytes, alveolates, stra- levels of prymnesiophyte-indicative pigments in regions where menopiles, and prasinophytes. Contributions by individual taxa, most Chl a (representing all phytoplankton combined) is in the however, are not well known, and have been analyzed <2-μm size fraction (6, 12). Six picoplanktonic prymnesiophytes from only the latter two lineages. Tiny “picoplanktonic” members of exist in culture (6, 7) but prymnesiophyte 18S rDNA sequences the prymnesiophyte lineage have long been inferred to be ecolog- from <2–3-μm size-fractioned environmental samples typically ically important but remain poorly characterized. Here, we examine belong to uncultured taxa (6, 13–15). As a whole, this lineage pico-prymnesiophyte evolutionary history and ecology using culti- reportedly diverged from other major eukaryotic lineages early vation-independent methods. 18S rRNA analysis showed pico- on, 1.2 billion years ago (16), and their overall placement among prymnesiophytes belonged to broadly distributed uncultivated eukaryotes is uncertain (4, 16). They are extremely distant from taxa. Therefore, we used targeted metagenomics to analyze uncul- phytoplankton with published genomes. Thus, although infer- tured pico-prymnesiophytes sorted by flow cytometry from sub- ences exist regarding their importance and evolutionary history, tropical North Atlantic waters. The data reveal a composite uncertainties surround even the most basic features of oceanic nuclear-encoded gene repertoire with strong green-lineage affilia- pico-prymnesiophytes, such as cell size, , growth rates, and tions, which contrasts with the evolutionary history indicated by the genomic composition. genome. Measured pico-prymnesiophyte growth rates were One approach for gaining insights to uncultivated taxa is meta- rapid in this region, resulting in primary production contributions . However, unicellular eukaryotes have larger genomes Prochlorococcus and lower gene density than marine and archaea and are similar to the cyanobacterium . On average, pico- fi fi prymnesiophytes formed 25% of global picophytoplankton bio- less abundant, making ef cient sequence recovery dif cult by seawater filtration. Parsing of eukaryotic data from diverse com- mass, with differing contributions in five biogeographical provinces munities is particularly problematic owing to the paucity of relevant spanning tropical to subpolar systems. Elements likely contributing reference genomes. Selection of DNA from an uncultivated target to success include high gene density and potentially involved microbe(s) (e.g., by fosmid sequencing or using cells sorted by flow in defense and nutrient uptake. Our findings have implications cytometry) obviates bioinformatic parsing issues and has revealed reaching beyond pico-prymnesiophytes, to the prasinophytes and unique gene complements in uncultured prokaryotes (17, 18). stramenopiles. For example, prevalence of putative Ni-containing To address uncertainties regarding pico-prymnesiophyte ecol- superoxide dismutases (SODs), instead of Fe-containing SODs, ogy, we integrated a suite of cultivation-independent methods. seems to be a common adaptation among eukaryotic phytoplank- Targeted metagenomics was developed to investigate diversity ton for reducing Fe quotas in low-Fe modern . Moreover, and genomic features of uncultivated pico-prymnesiophytes. highly mosaic gene repertoires, although compositionally distinct Growth rates were measured and used to assess primary pro- for each major eukaryotic lineage, now seem to be an underlying duction in the same region. Building on this contextual dataset, facet of successful marine phytoplankton. biomass contributions were evaluated across provinces spanning tropical to subpolar systems, providing a comprehensive view of comparative genomics | primary production | prymnesiophytes | marine global importance and latitudinal variations. photosynthesis | haptophytes

lobal primary production is partitioned equally among terres- Author contributions: M.L.C., T.I., R.S.L., and A.Z.W. designed research; M.L.C., A.M., Gtrial and marine ecosystems, each accounting for ≈50 gigatons S.G.T., T.W., R.M.W., T.I., B.J.B., M.L., C.G., K.R.B., J.H., and A.Z.W. performed research; B.R. of carbon per year (1). The phytoplankton responsible for marine contributed new reagents/analytic tools; M.L.C., A.E.A., A.M., J.P.M., M.M., S.G.T., T.W., J.-H.L., C.L.D., M.L., M.T., E.C., F.P.C., and A.Z.W. analyzed data; and M.L.C., A.E.A., J.P.M., primary production include the , Prochlorococcus and S.G.T., T.W., and A.Z.W. wrote the paper. , and a multitude of eukaryotic phytoplankton, such as The authors declare no conflict of interest. diatoms, dinoflagellates, prasinophytes, and prymnesiophytes (2–4). “ ” < – μ Data deposition: The sequences reported in this paper have been deposited in the Gen- Most oceanic phytoplankton are picoplanktonic ( 2 3 mdi- Bank database (accession nos. HM581528–HM581638 and HM565909–HM565914). Other ameter) and have high surface area to volume ratios, an advantage in scaffolds with predicted genes from this Whole Genome Shotgun/454 project have been open- low-nutrient conditions (5–8). Despite the importance deposited at DNA Data Bank of Japan/European Molecular Biology Laboratory/GenBank of eukaryotic phytoplankton to carbon cycling only six genomes have under the accession no. AEAR00000000. The version described in this paper is the first been sequenced and analyzed comparatively, all being from diatoms version, AEAR01000000. and prasinophytes. These revealed greater differentiation than an- This article is a PNAS Direct Submission. ticipated on the basis of 18S rRNA gene analyses (9–11). The ob- Freely available online through the PNAS open access option. served genomic divergence is associated with major differences in 1A.E.A. and A.M. contributed equally to this work. 2 physiology and niche adaptation (10). To whom correspondence should be addressed. E-mail: [email protected]. SCIENCES

Pigment-based estimates indicate that prymnesiophytes, also This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. ENVIRONMENTAL known as haptophytes, are broadly distributed and abundant. 1073/pnas.1001665107/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1001665107 PNAS | August 17, 2010 | vol. 107 | no. 33 | 14679–14684 Downloaded by guest on October 1, 2021 Results and Discussion and Sanger technologies (SI Materials and Methods, Section 5). -Targeted Metagenomics Approach and Diversity Context. Genes were modeled on assembled scaffolds and then screened Photosynthetic picoeukaryote populations were sorted by flow phylogenetically using the E. huxleyi nuclear genome. For selection, fi cytometry (hereafter “sorted” or “the sort”) based on scatter and half the identi able genes on a scaffold had to clade directly with autofluorescence characteristics from two subtropical North At- E. huxleyi, to the exclusion of gene sequences from all other taxa (SI lantic samples collected in the Florida Straits (Fig. 1, Fig. 2 Inset, Materials and Methods, Section 6). This ensured that only prymne- and SI Materials and Methods, Sections 1 and 2). Whole-genome siophyte-derived scaffolds were further analyzed, because the stra- amplification (19) was performed on sorted target populations (SI menopile Pelagomonas fell partially in the same flow-cytometric Materials and Methods, Section 3), providing unprecedented access population (SI Materials and Methods, Section 5). Seventy-one per- to pico-prymnesiophyte DNA. cent of genes on selected scaffolds were sistered by E. huxleyi genes, The sorted pico-prymnesiophytes were distinct from cultured demonstrating screening rigor; only this scaffold subset (2 MB of taxa but closely related to environmental sequences from native assembly) was considered unambiguously pico-prymnesiophyte populations. 18S rDNA clone libraries built from the sort DNA derived and used for comparative analyses. Twenty-nine percent of were analyzed within the context of <2–3 μm size-fractionated genes on pico-prymnesiophyte scaffolds seemed to be missing from clone libraries from the Florida Straits, the Sargasso Sea, the E. huxleyi, supporting their distance from coccolithophores. Pacific Ocean (Fig. 2, SI Materials and Methods, Section 4, Fig. S1A, and Table S1), and published data. Similar to other studies Gene Density and Genome Size Predictions. Prymnesiophytes are (6, 13, 15), the vast majority of sequences were from uncultured currently placed in the controversial eukaryotic Supergroup Chro- prymnesiophytes. The majority of prymnesiophyte sequences malveolata along with stramenopiles (e.g., diatoms) and alveolates from the Florida Straits Station 04 sort belonged to environ- (e.g., dinoflagellates) but separate from the Supergroup Arch- mental group 8 (111 clones, 99–100% identity; Fig. S1B), also aeplastida, which includes prasinophytes and all green-lineage present in the easterly Station 08 sort. Lower levels of group 3 organisms (e.g., land plants). However, their genome characteristics were detected in the Station 04 sort (7 clones, 99–100% identity), are unknown and phylogenetic placement uncertain. We found along with sequences at the tip of the tree (e.g., group 15; SI high gene density in the pico-prymnesiophyte metagenome (Fig. 3A Materials and Methods, Section 5). Group 8 was also seen in the and SI Materials and Methods, Section 7), akin to pico-prasino- Sargasso Sea (14) and on multiple dates in the Florida Straits phytes, which purportedly underwent genome streamlining to op- (Table S2). Group 8 had only 93% 18S rDNA identity to any timize life in oligotrophic niches (9, 11). GC content was 60%. cultured organism (to Chrysochromulina), and phylogenetic place- Genome size was calculated using multiple methods accommo- ment was unresolved (no bootstrap support). dating the composite of the sorted pico-prymnesiophyte The observed level of group 8 18S rDNA divergence has im- population (SI Materials and Methods, Section 7). The first involved portant implications for gene content. The diatoms Thalassiosira a model linking the size distribution of euKaryotic Orthologous pseudonana and Phaeodactylum tricornutum, with 90% 18S Groups of proteins (KOG), allocation of functions across those rDNA identity, share only 30–40% of their genes and occupy families and total gene content for 12 protistan reference genomes. fundamentally different niches (10). Of the four picoeukaryote This rendered an average genome content of 12,711 (±1,145) genes genomes (all prasinophytes), two Micromonas isolates, with 97% (Fig. S2 A and B), such that the 1,624 pico-prymnesiophyte nuclear- 18S rDNA identity, have 69.5% DNA identity over aligned ge- encoded genes identified constituted 13% of those in a nome regions, sharing at most 90% of their protein-encoding representative genome. Results were similar (11,600 total genes; genes (11). Although an unpublished genome of the coccoli- SI Materials and Methods, Section 7) using a method based on counts thophore Emiliania huxleyi is available, this prymnesiophyte is of 132 near single-copy core genes in other eukaryotes. Diatom ge- larger than the pico-prymnesiophytes and expected to occupy nome content is comparable, ≈10,000–14,000 genes (10), whereas a distinct niche (8, 16). Furthermore, the soft-bodied prymne- the pico-prasinophytes contain 8,000 to just over 10,000 genes (9, 11). siophytes in the sort were distant from coccolithophores (Fig. 2), a group known for their calcium carbonate plates (16). Mosaic Gene Repertoires. Evolutionary and functional aspects of Given the differences between the group 8–enriched sort and nuclear genomes were analyzed using gene con- population and cultured taxa, we used targeted metagenomics to tent. 16S rRNA genes revealed pico-prymnesiophyte chloroplast discover genomic features of environmentally relevant pico- genome scaffolds; the largest (scaffold C19847), containing 45 prymnesiophytes. Station 04 sort DNA was sequenced by 454-FLX protein-encoding genes among others (Tables S3 and S4), seemed to be from group 8 (SI Materials and Methods, Sections 5 and 6, and Figs. S1B and S3A). Phylogenetic analysis of 22 concatenated 4

10 plastid genome-encoded genes conserved across all lineages Sorted Eukaryotes placed this pico-prymnesiophyte within the Chromalveolata, be- Eukaryotes tween cryptophytes and stramenopiles, and directly sistered by

3 E. huxleyi (SI Materials and Methods, Section 7, and Fig. S3B), the 10 Synechococcus only published prymnesiophyte plastid genome (20). This place- ment corresponded to that for E. huxleyi in other plastid phylog- enies (21), but with greater support. Nucleotide level comparison 2

10 of C19847 with E. huxleyi showed large-scale rearrangements akin to the divergent diatoms (Fig. S3C). Gene-level synteny was rel- Prochlorococcus atively conserved with E. huxleyi (Fig. S3D).

1 In contrast, global analysis of the 1,624 predicted pico-

Red fluorescence Bead 10 position prymnesiophyte nuclear-encoded genes revealed a distinctive evo- lutionary signature relative to plastid genome-encoded elements. Phylogenomic analyses of nuclear-encoded genes indicated that these phytoplankton are evolutionarily situated between the green- 101 102 103 104 lineage and stramenopiles, with greater apparent affinity to the Forward angle light scatter former. The gene repertoire shared a higher degree of overlap with Fig. 1. Forward angle light scatter and chlorophyll-derived red fluorescence prasinophytes (Fig. 3B). Even among pico-prymnesiophyte genes characteristics of picophytoplankton groups (at Florida Straits Station 04; with homologs in both prasinophytes and stramenopiles, similarity Fig. 2 Inset, blue arrow), including the metagenome population (magenta was generally higher to prasinophytes. Of 352 phylogenetic trees circle) and showing the position of 0.75-μm bead standards (black box, not constructed herein that contained at least one green-lineage or- added to actual sort to avoid contamination). ganism and one stramenopile, the genetic distance between the

14680 | www.pnas.org/cgi/doi/10.1073/pnas.1001665107 Cuvelier et al. Downloaded by guest on October 1, 2021 Pavlovales Temp. Depth Geo.loca. 1 2

3 4 Phaeocystis jahnii

Phaeocystis antartica SK2 Phaeocystis pouchetii (Hariot) Lagerheim Phaeocystis pouchetii P360 Phaeocystis globosa 6 Temperature (oC) Depth (m) Coccolithophores 8 5.0-9.9 9 0-49 10 Braarudosphaera bigelowii TP05-6-b 10.0-14.9 50-99 Chrysochromulina parkeae Braarudosphaera bigelowii 15.0-19.9 TP05-6-a 100-150 Imantonia 20.0-24.9 3499 Chrysochromulina hirta Chrysochromulina herdlensis 25.0-29.9 Chrysochromulina fragaria Chrysochromulina brevefilium >30.0 Prymnesium sp. UIO 133 Chrysochromulina kappa

Chrysochromulina minor Chrysochromulina polylepis Geographic location 11 Prymnesium 12 13 Florida Chrysochromulina scutellum Prymsiophyte symbiont 1 USA Chrysochromulina acantha

FL Straits FL Chrysochromulina throndsenii Chrysochromulina cymbium Bahamas Chrysochromulina campanulifera Chrysochromulina sp. MBIC10513 Chrysochromulina sp. UIO TH2 Chrysochromulina rotalis Chrysochromulina parva

14 BATS Chrysochromulina leadbeateri Sargasso Sea

15 16 17 18 Chrysochromulina sp. NIES-1333 Chrysochromulina simplex 19 20 0.1 21 22

Fig. 2. Maximum-likelihood reconstruction of (blue lines) environmental and (black) cultured prymnesiophyte 18S rDNA sequences from two sorts and 25 size- fractionated samples from discrete depths, dates, and locations (circles; Table S1) and previous publications (triangles) (SI Materials and Methods, Section 4). The Station 04 sort (blue arrow) was in the Gulf Stream core. A single representative was used for redundant sequences within each library. Supported clades composed of only environmental sequences were collapsed after tree building when of ≥99% identity (blue groups 1–22; Table S2). Node support is shown for (black circles) 100% and (black triangles) ≥75% support. Uncultured prymnesiophytes have also been seen recently in the South Pacific(40).

pico-prymnesiophyte and the former was shorter twice as often (Fig. The discontinuity in gene content could help explain puzzling S4A). Across all trees, the mean relative genetic distance was 6.5% ambiguities in prymnesiophyte evolution. The chloroplast ge- shorter between pico-prymnesiophytes and Archaeplastida, com- nome would drive the relationship toward a red-lineage sec- pared with stramenopiles, calling into question the current, largely ondary endosymbiosis (Chromalveolata; Fig. S3B), whereas the plastid genome-based, phylogenetic placement of this lineage. nuclear genome retains features of a green host (Fig. 3B and Fig. More than 75% of pico-prymnesiophyte genes assigned to KOGs S4A). Of pico-prymnesiophyte genes that appeared more similar were in the largest KOG families (top 20%), whereas for other to Archaeplastida than to stramenopiles, 55% were closer to marine phytoplankton this percentage was lower (SI Materials and streptophytes, particularly early diverging plants, suggesting Methods, Section 7). This high level of KOG redundancy reflected a strong ancestral green-lineage influence in the prymnesiophyte the mosaic nature of pico-prymnesiophyte gene repertoires. The host organism’s gene pool; only 45% were closer to green functional KOG repertoire straddled prasinophytes and strameno- (Fig. S4B). Alternatively, like diatoms (22), prymnesiophytes piles, such that several expansions in the metagenome seemed to may have obtained green-lineage genes from an ancient cryptic be absent from one or both of these other lineages. Presence of endosymbiont. Although the paucity of eukaryotic phytoplank- more than one pico-prymnesiophyte taxon in the sort does not ex- ton genomes may influence results, nuclear-encoded marker plain this redundancy, because it would require disproportionate genes from larger cultured prymnesiophytes also show strong sampling of the same KOGs, to the exclusion of others, from each green-lineage affiliations (3). We anticipate that E. huxleyi ge- pico-prymnesiophyte genome. Families with multiple representa- nome analysis will confirm that the mosaic gene repertoire tives included nudix hydrolases and arylsulfatases, not found in reported here is a lineage-wide characteristic. prasinophytes, but present in metazoa and bacteria (arylsulfatases

are also in stramenopiles) (Table S5). Polyketide synthases were Functional Gene Repertoire. Pico-prymnesiophyte nuclear-encoded SCIENCES

also found and present in prasinophytes and bacteria but missing genes also showed differences in functional composition (Fig. ENVIRONMENTAL from diatoms. 3C). Although all transcription factor families recovered were

Cuvelier et al. PNAS | August 17, 2010 | vol. 107 | no. 33 | 14681 Downloaded by guest on October 1, 2021 Fig. 3. Characteristics of the pico-prymnesiophyte metagenome. (A) Gene density histograms as the proportion of nucleotides in genes over all chromosomes within an incrementally sampled sliding window in prasinophytes and photosynthetic (diatoms and Aureococcus) and nonphotosynthetic (Phytophthora)stra- menopiles. Average gene density on pico-prymnesiophyte scaffolds was 74% (magenta line). (B) Venn diagram of pico-prymnesiophyte genes in relation to prasinophyte (green, Micromonas and Ostreococcus), diatom (beige), and Phytophthora (yellow) genes by BLASTP (e-value ≤1.0e−9). Numbers indicate gene counts in each Venn group (magenta letters), whereas pie charts show relative best BLAST proportions for overlapping groups. Some pico-prymnesiophyte genes (blue text) were not found in the other lineages. (C) Comparison of functional profiles of pico-prymnesiophyte–specific genes with those from each Venn group by high- level Gene Ontology (GO) molecular functions. Relative proportions are shown for each GO function within a Venn group such that each column represents100%.

also in Archaeplastida and stramenopiles, higher proportions of Environmental Importance. Increased stratification and lower nutri- specific regulators were observed, such as chromatin remodeling ent concentrations predicted under some climate-change scenarios genes (Tables S5 and S6), which play roles in meiosis and “life- are hypothesized to create conditions favoring picophytoplankton style” changes. These included divergent SET-domain proteins over larger species (11). However, questions remain regarding the and SNF2/helicases as chromatin remodeling factors (23). Dis- ecological importance of pico-prymnesiophytes in today’s condi- tinct SUV39 subfamily members (SET) identified were not found tions, against which physiological stressors and future changes can in stramenopiles (Fig. S5A). This family is responsible for het- be assessed. To establish their roles, we developed a contextual erochromatin formation (23) and could enable control over in- metadataset for the pico-prymnesiophyte metagenome, including vasive entities like transposons and viruses. distributions at the subtropical North Atlantic sites (Fig. 2, Inset) Other features corresponded to life in modern oceans, for ex- by FISH. Biomass contributions were compared with the overall ample, a Ni-containing superoxide dismutase (Ni-SOD). SODs are picophytoplankton community, specifically, other picoeukaryotes, vital to photosynthetic organisms, scavenging toxic superoxide rad- Synechococcus and Prochlorococcus (SI Materials and Methods, icals generated by multiple metabolic pathways including photo- Sections 2, 9, and 10). synthesis. Isoforms use different metals at the active site, influencing Pico-prymnesiophyte contributions in the Sargasso Sea were trace metal requirements (24). Presence of sodN, which encodes Ni- roughly equivalent at the surface and deep chlorophyll maximum SOD, corresponds with absence of Fe-SOD–encoding genes in (DCM), representing 23% and 21% of picophytoplankton carbon, many marine bacterial genomes, and Ni-SOD seems to be more respectively. Two pico-prymnesiophyte size classes were evident, prevalent in open-ocean strains (25). This results in replacement of cells of 1.9 ± 0.4 × 2.1 ± 0.3 μm(n = 89) and 2.8 ± 0.6 × 3.4 ± 0.5 some cellular Fe demand with Ni, a valuable adaptation considering μm(n =127)(Table S7). In the Florida Straits their contributions 10-fold higher Ni concentrations in oceanic waters (25). We iden- were typically higher in surface waters, although more nutrients tified putative sodN genes in the pico-prymnesiophyte meta- were presumably available at the DCM (Fig. S6A). Here, 90% ± genome, the four pico-prasinophyte genomes (see also ref. 9) and P. 9% and 87% ± 13% of prymnesiophytes were <3 μm, averaged tricornutum (Fig. S5B). T. pseudonana did not seem to encode sodN, over 1 y, at the surface and DCM, respectively. The direct-count– but rather a Fe-SOD, correlating with its coastal distribution and based biomass approach resulted in lower pico-prymnesiophyte higher iron quotas than P. tricornutum (26). Fe-SOD–encoding DCM contributions than estimated by HPLC (Fig. S6B). However, genes were not found in the sodN-containing phytoplankton these methods gave similar surface trends, supporting the HPLC- genomes. Presence of putative sodN genes in oceanic phytoplank- based inference that picoplanktonic taxa form most prymnesio- ton from three major eukaryotic lineages suggests that replacement phyte biomass in open-ocean surface waters (6, 12, 28). of Fe-SOD with Ni-SOD may be a common adaptation to the lower Pico-prymnesiophyte specific growth rates showed that these tiny availability of iron in modern oceans than in past times (27). eukaryotes can grow rapidly, amplifying contributions to primary Several domains involved in uptake of large substrates, such as production. Rates were measured by the dilution method and di- proteins and nucleic acids, as well as salvage of nucleosides, were rect counts (SI Materials and Methods, Section 11), which render represented more highly in the pico-prymnesiophyte meta- growth rates close to in situ cell cycle–based rates for taxa amenable genome (Fig. 3C). These could facilitate uptake of otherwise in- to the latter analysis (29, 30). Pico-prymnesiophyte growth rates − accessible nutrients. For example, a member of the amino acid/ in the Sargasso Sea were high at 15 m (1.12 d 1, r2 = 0.87, P < − polyamine/organocation superfamily is a likely transporter for the 0.07) and lower at 70 m (0.29 d 1, r2 = 0.73, P < 0.07). Pro- − nitrogen-rich guanine derivative xanthine, potentially important chlorococcus grew more slowly (0.63 d 1, r2 = 0.54, P < 0.01) than − under nitrogen-limiting conditions. A plant-like putative acid pico-prymnesiophytes at the surface, but faster (0.60 d 1, r2 = phosphatase (Table S5), which likely cleaves intracellular ortho- 0.61, P < 0.001) at depth. Because the pico-prymnesiophyte data phosphoric-monoesters to phosphate, could also be involved in constitute the first specific growth rates reported, for experimental nutrient scavenging, although this role is speculative. Overall, validation, we compared Prochlorococcus growth rates with pre- features of the metagenome suggest adaptations associated with vious direct count-based rates from the same region and time of − survival in oligotrophic environments. year, which were similar (0.52 d 1) (30).

14682 | www.pnas.org/cgi/doi/10.1073/pnas.1001665107 Cuvelier et al. Downloaded by guest on October 1, 2021 Table 1. Average surface biomass of picophytoplankton groups in five biogeographical provinces − Biomass (μgCL 1)

Ocean area “Non-prym” Latitude (× 1012 m2) Samples (n) Prochlorococcus Synechococcus picoeukaryotes Pico-prymnesiophytes

60°–45°N 13.22 8 0.1 (0.1) 2.4 (1.1) 3.0 (1.7) 7.0 (4.7) 45°–20°N 47.18 24 3.7 (3.0) 0.5 (0.5) 1.0 (2.7) 2.0 (1.5) 20°N–20°S 122.60 59 8.4 (2.8) 1.0 (1.1) 1.8 (1.7) 1.8 (0.9) 20°–45°S 75.35 19 3.1 (2.0) 0.4 (0.6) 3.7 (1.4) 1.8 (0.9) 45°–65°S 48.69 11 0.3 (0.4) 1.1 (1.9) 3.9 (2.8) 5.5 (4.9)

Area varies over latitudinal zones owing to the influence of land masses; sample number is also shown, with values from sites sampled seasonally averaged and counted here as 1 sample (Table S8). Values in parenthesis reflect SDs.

We combined biomass with specific growth and mortality rates to pigments and their contribution to total Chl (15). Validation (of estimate primary production. Despite lower pico-prymnesiophyte ref. 15) would result in major reevaluation of tropical systems abundance, their combined greater cellular biomass and faster where the streamlined genome, small size, and low nutrient quotas growth led to primary production levels comparable to Prochloro- of Prochlorococcus seem highly advantageous given extended pe- − − coccus at 15 m (1.1 and 1.8 μgCL 1 d 1, respectively; SI Materials riods of stratification (see, e.g., ref. 17). The significant discrepancy and Methods, Section 11). Pico-prymnesiophyte primary production with our results may stem from issues surrounding the algorithm- was almost 4-fold higher than that of all other picoeukaryotes based approach used in ref. 15, such as (i) the fact that other lin- − − (0.27 μgCL1 d 1). Total primary production at the Bermuda eages can contain the prymnesiophyte-indicative marker pigment Atlantic Time-series Study (BATS; Fig. 2, Inset)inmid-May,be- (28, 33) (SI Materials and Methods, Section 8), (ii) a variable re- fore our experiments, and just afterward, in mid-June, was 7.81 and fi − − lationship between a speci c pigment content and surface Chl, or 0.96 μgCL 1 d 1, respectively, at 20 m (31). Seasonal stratification (iii) not partitioning contributions by organism size and the fact developed during this period, leading to lower production by mid- that HPLC samples are not size-fractionated. Furthermore, our June. Over a decade, BATS total primary production occasionally – − − − tropical surface direct-count based Prochlorococcus biomass data ranged up to 6 μgCL 1 d 1 at 20 m, but was typically 2–4 μgCL 1 − corresponded well with that from HPLC (28, 34), which, for Pro- d 1 each June; the pico-prymnesiophyte value constitutes 25–50% chlorococcus, is less prone to such caveats. of that. Likewise, a flow cytometrically defined picoeukaryote Pico-prymnesiophytes seem to be highly successful and show population (73% ± 2% prymnesiophyte cells, over 20 stations), was signs of optimization to open-ocean conditions. Features within recently shown to perform 25% ± 9% of primary production in the the metagenome remain difficult to relate to niche differentia- North East Atlantic, ranging up to 38% (32). Plastid-16S rRNA tion given 37% genes of unknown function, similar to many genes were evaluated at one station in that study and showed un- genomes. This lack of functional understanding is perhaps the cultured prymnesiophyte taxa, as found herein. greatest impediment to connecting genomes to organism physi- ology and response. The research herein positions us to explore Global Contributions and Latitudinal Gradients. Our results, taken the function of such genes in situ. with those from the North East Atlantic (32), point to significant pico-prymnesiophyte contributions in the subtropical North At- Conclusions lantic. However, variations between oceans and latitudinal gra- Soft-bodied prymnesiophytes survived the K/T boundary mass dients translate to major biotic differences, and the respective extinctions (16), indicating that taxa akin to those analyzed herein communities will likely respond to change differently. To assess were resilient to perturbation. Surface water warming has now global surface biomass contributions by pico-prymnesiophytes we been correlated with increases of picophytoplankton in the Arctic fi fi counted and sized cells in ve biogeographical provinces, speci - Ocean (35) and will presumably impact lower latitudes. However, cally subpolar (high-latitudes) and subtropical-temperate (mid- the success of small prymnesiophytes and their contributions in latitudes) systems, as well as the tropics (low latitudes) (SI Materials future times are linked to evolutionary history and genetic makeup, and Methods, Sections 1, 2, 9, and 10,andTable S8). as well as the rate and extent of perturbations experienced. ± μ −1 Globally, pico-prymnesiophytes averaged 2.6 1.8 gCL Our results, showing genomic features of pico-prymnesiophytes, when the areal extent of each province was accounted for (Table 1). their rapid growth and significant global contributions, provide This amounted to ≈50% of that of Prochlorococcus (4.7 ± 2.1 μgC − L 1), which was less evenly distributed and nearly absent from cold waters. The considerable biomass of pico-prymnesiophytes was o again in part due to larger cell size than other picophytoplankton 80 N 30 (6), and their contributions were less obvious by abundance alone (Table S8); Prochlorococcus, for instance, is orders of magnitude 60 oN 25 more abundant in low- and mid-latitudes but is also much smaller. 40 oN 20 Biomass showed a strong latitudinal signal. In high latitudes, o pico-prymnesiophytes dominated, comprising 50–56% of pico- 20 N phytoplankton biomass (Fig. 4, Table 1, and Table S8). Relative 0 15 contributions in mid-latitudes were modified by variations in other 20 oS groups. Pico-prymnesiophyte biomass per liter was maximal in the 40 oS 10 northern subpolar province, but the massive extent of the Southern Ocean, relatively unimpeded by land, rendered their greatest 60 oS 5 contributions to global biomass in the southern subpolar province. o 100% 50% 20% Pico-prymnesiophyte contributions were lowest in the tropics 80 S o o o o o 0 − 180 W 120 W 60 W 0 60 E 120 E oC (1.8 μgCL 1), comprising approximately one fifth (21%) of Pro- chlorococcus biomass (Table 1). These in situ observations were at Fig. 4. Global surface biomass contributions of prymnesiophytes as percent-

odds with a recent report suggesting that pico-prymnesiophytes age of total picophytoplankton carbon, represented by bubble size (scaling at SCIENCES

are more important than Prochlorococcus in low latitudes, based lower right). Sea surface temperature represents 1° increments averaged ENVIRONMENTAL on an algorithm relating satellite surface Chl to prymnesiophyte monthly over 18 y; note differences in the five biogeographical provinces.

Cuvelier et al. PNAS | August 17, 2010 | vol. 107 | no. 33 | 14683 Downloaded by guest on October 1, 2021 several key advancements. Although pico-prymnesiophytes are by phylogenomic methods using predicted genes and a database containing, clearly diverse, aspects of composition, potential streamlining, and among others, 46 eukaryotic genomes including 10 phytoplankton and 2 functional attributes revealed by targeted metagenomics provide additional protistan stramenopile genomes. Comparative gene analyses with insights into their evolutionary success. Together with data on E. huxleyi are prohibited and were not performed. Pico-prymnesiophyte prasinophytes and stramenopiles (10, 11), our analysis of the pico- predicted proteins were characterized by profiling and genome-size estimate prymnesiophyte metagenome indicates that mixed-lineage gene methods. Chloroplast scaffolds were manually selected and annotated. repertoires are a transcendent property of successful phytoplank- Microscopy was used with a prymnesiophyte-specific FISH probe (36) to ton in modern oceans, not a rare feature. These mosaic gene count and size cells. For some cruises identification was by chloroplast ar- repertoires seem to be compositionally distinct for each lineage, rangement, flagellar characteristics, and occasional presence of a hapto- fl in uenced by disparate sources of the tree of life, or differential nema; no significant difference (t test, P = 0.43) was detected for Atlantic data gene loss from an ancestor, and presumably factors driving niche from 25° to 35° N (FISH: 500 ± 61 mL−1, n = 26; characteristics-based: 593 ± 108 differentiation. Furthermore, the complexity of marine microbial −1 fl fi mL , n = 12). Other picophytoplankton groups were enumerated by ow communities makes it dif cult to determine taxa critical to eco- cytometry (37). Biovolume was calculated from cell size for pico-prymnesio- system processes and global biogeochemical models; our work phytes and biomass determined using an established biovolume-based car- highlights the need to prioritize pico-prymnesiophytes. Finally, it opens the door for research on the physiology and response bon conversion factor, also used for other picophytoplankton groups (38) (Table S7). Dilution experiments were according to (28, 30) and counts by FISH capabilities of uncultivated members of this ancient primary fl producer lineage. and ow cytometry. HPLC data were analyzed according to ref. 39. Materials and Methods ACKNOWLEDGMENTS. We thank the captains and crews of research vessels Discoverer, Endeavor, Ka’imimoana, Malcolm Baldridge, Oceanus, Walton Methods details are in SI Materials and Methods. Samples were from the Smith, and Western Flyer; cruise participants, especially F. Not; J. Heidelberg, North, Equatorial, and South Atlantic, the North, Equatorial, and West Pa- R. Gausling, and G. Weinstock for 18S rDNA sequencing; M. Kogut, S. Giovan- cific, and the Indian Ocean(s). 18S rDNA clone libraries (14) were constructed noni, and R. Gausling for edits; and J. Eisen. Sequencing was under DE-AC02- from subtropical Atlantic and North West Pacific size-fractionated samples, 05CH11231, by a Department of Energy Community Sequencing Program or from cells sorted by flow cytometry. award to A.Z.W. and J. Eisen. Support was in part by DE-FC02-02ER63453, Approximately 300 cells from target populations in the subtropical North NSF OCE-0722374, and NSF-MCB-0732448 (to A.E.A.); a National Human Genomic Research Institute, National Institutes of Health grant (to R.S.L.); Atlantic were sorted by flow cytometry. DNA from two populations was fi fi National Oceanic and Atmospheric Administration and David and Lucile Pack- ampli ed by multiple displacement ampli cation (19) and used for PCR-based ard Foundation (DLPF) grants (F.P.C.); NSF-OCE-0241740 (to B.J.B.); and major clone libraries and, for one sample, metagenomic sequencing. Metagenomic funding by NSF-OCE-0836721, the DLPF, and a Moore Foundation Young In- data were assembled in a two-step high-stringency process and only scaffolds vestigator Award as well as Moore 1668 (to A.Z.W.). Author contribution resulting from the second stage considered further. Scaffolds were screened details are given in SI Materials and Methods, Section 12.

1. Field CB, Behrenfeld MJ, Randerson JT, Falkowski P (1998) Primary production of the 23. Dillon SC, Zhang X, Trievel RC, Cheng XD The SET-domain protein superfamily: biosphere: Integrating terrestrial and oceanic components. Science 281:237–240. Protein lysine methyltransferases. Genome Biol 2005; 6: 227. 2. Chisholm SW, et al. (1992) Prochlorococcus marinus nov gen nov sp: An oxyphototrophic 24. Wolfe-Simon F, Starovoytov V, Reinfelder JR, Schofield O, Falkowski PG (2006) marine prokaryote containing divinyl chlorophyll a and b. Arch Microbiol 157:297–300. Localization and role of manganese superoxide dismutase in a marine diatom. Plant 3. Hampl V, et al. (2009) Phylogenomic analyses support the monophyly of Excavata and re- Physiol 142:1701–1709. solve relationships among eukaryotic “supergroups”. Proc Natl Acad Sci USA 106:3859–3864. 25. Dupont CL, Buck KN, Palenik B, Barbeau K (2010) Nickel utilization in phytoplankton 4. Keeling PJ, et al. (2005) The tree of eukaryotes. Trends Ecol Evol 20:670–676. assemblages from contrasting oceanic regimes. Deep-Sea Res Pt I 57:553–566. 5. Chisholm SW (1992) Primary Production and Biogeochemical Cycles in the Sea, eds 26. Kustka AB, Allen AE, Morel FMM (2006) Sequence analysis and transcriptional Falkowski PG, Woodhead AD (Plenum Press, New York), pp 213–237. regulation of iron aquisition genes in two marine diatoms. J Phycol 43:715–729. 6. Worden AZ, Not F (2008) of the Oceans, ed Kirchman DL (John 27. Falkowski PG, et al. (2004) The evolution of modern eukaryotic phytoplankton. Wiley & Sons, Hoboken, NJ), 2nd Ed, p 594. Science 305:354–360. 7. Vaulot D, Eikrem W, Viprey M, Moreau H (2008) The diversity of small eukaryotic 28. Landry MR, et al. (2003) Phytoplankton growth and microzooplankton grazing in phytoplankton (< or =3 microm) in marine ecosystems. FEMS Microbiol Rev 32:795–820. high-nutrient, low-chlorophyll waters of the equatorial Pacific: Community and 8. Raven JA (1998) Small is beautiful: the picophytoplankton. Funct Ecol 12:503–513. taxon-specific rate assessments from pigment and flow cytometric analyses. 9. Palenik B, et al. (2007) The tiny eukaryote Ostreococcus provides genomic insights J Geophys Res-Oceans 108:8142–8155. into the paradox of speciation. Proc Natl Acad Sci USA 104:7705–7710. 29. Landry MR (1993) Handbook of Methods in Aquatic Microbial Ecology, eds Kemp PF, 10. Bowler C, Vardi A, Allen AE (2010) Oceanographic and biogeochemical insights from Sherr BF, Sherr EB, Cole JJ (Lewis Publishers, Boca Raton, FL), pp 715–722. diatom genomes. Ann Rev Mar Sci 2:429–461. 30. Worden AZ, Binder BJ (2003) Application of dilution experiments for measuring 11. Worden AZ, et al. (2009) Green evolution and dynamic adaptations revealed by growth and mortality rates among Prochlorococcus and Synechococcus populations in genomes of the marine picoeukaryotes Micromonas. Science 324:268–272. oligotrophic environments. Aquat Microb Ecol 30:159–174. 12. Mackey D, Blanchot J, Higgins H, Neveux J (2002) Phytoplankton abundances and 31. Lomas MW, et al. (2010) Increased ocean carbon export in the Sargasso Sea linked to community structure in the equatorial Pacific. Deep-Sea Res Pt I 49:2561–2582. climate variability is countered by its enhanced mesopelagic attenuation. 13. Moon-van der Staay S, et al. (2000) Abundance and diversity of prymnesiophytes in Biogeosciences 7:57–70.

the picoplankton community from the equatorial Pacific Ocean inferred from 18S 32. Jardillier L, Zubkov MV, Pearman J, Scanlan DJ (2010) Significant CO2 fixation by small rDNA sequences. Limnol Oceanogr 45:98–109. prymnesiophytes in the subtropical and tropical northeast Atlantic Ocean. ISME J, doi: 14. Not F, Gausling R, Azam F, Heidelberg JF, Worden AZ (2007) Vertical distribution of 10.1038/ismej.2010.36. picoeukaryotic diversity in the Sargasso Sea. Environ Microbiol 9:1233–1252. 33. Carreto JI, Seguel M, Montoya NG, Clement A, Carignan MO (2001) Pigment of the 15. Liu H, et al. (2009) Extreme diversity in noncalcifying haptophytes explains a major ichthyotoxic dinoflagellate Gymnodinium sp. from a massive bloom in southern Chile. pigment paradox in open oceans. Proc Natl Acad Sci USA 106:12803–12808. J Plankton Res 23:1171–1175. 16. Medlin L, Saez AG, Young JR (2008) A molecular clock for coccolithophores and 34. Letelier RM, et al. (1993) Temporal variability of phytoplankton community structure implications for selectivity of phytoplankton extinctions across the K/T boundary. Mar based on pigment analysis. Limnol Oceanogr 38:1420–1437. Micropaleontol 67:69–86. 35. Li WKW, McLaughlin FA, Lovejoy C, Carmack EC (2009) Smallest algae thrive as the 17. DeLong EF, Karl DM (2005) Genomic perspectives in microbial oceanography. Nature Arctic Ocean freshens. Science 326:539. 437:336–342. 36. Simon N, et al. (2000) Oligonucleotide probes for the identification of three algal groups

18. Zehr JP, et al. (2008) Globally distributed uncultivated oceanic N2-fixing cyanobacteria by dot blot and fluorescent whole-cell hybridization. J Eukaryot Microbiol 47:76–84. lack oxygenic photosystem II. Science 322:1110–1112. 37. Buck KR, Chavez FP, Campbell L (1996) Basin-wide distributions of living carbon 19. Dean FB, et al. (2002) Comprehensive human genome amplification using multiple components and the inverted trophic pyramid of the central gyre of the North displacement amplification. Proc Natl Acad Sci USA 99:5261–5266. Atlantic Ocean, summer 1993. Aquat Microb Ecol 10:283–298. 20. Sánchez Puerta MV, Bachvaroff TR, Delwiche CF (2005) The complete plastid genome 38. Worden AZ, Nolan JK, Palenik B (2004) Assessing the dynamics and ecology of marine sequence of the haptophyte Emiliania huxleyi: a comparison to other plastid picophytoplankton: The importance of the eukaryotic component. Limnol Oceanogr genomes. DNA Res 12:151–156. 49:168–179. 21. Le Corguillé G, et al. (2009) Plastid genomes of two brown algae, Ectocarpus 39. Wright SW, et al. (2009) Composition and significance of picophytoplankton in siliculosus and Fucus vesiculosus: further insights on the evolution of red-algal derived Antarctic waters. Polar Biol 32:797–808. . BMC Evol Biol 9:253. 40. Shi XL, Marie D, Jardillier L, Scanlan DJ, Vaulot D (2009) Groups without cultured 22. Moustafa A, et al. (2009) Genomic footprints of a cryptic plastid endosymbiosis in representatives dominate eukaryotic picophytoplankton in the oligotrophic South diatoms. Science 324:1724–1726. East Pacific Ocean. PLoS ONE 4:e7657.

14684 | www.pnas.org/cgi/doi/10.1073/pnas.1001665107 Cuvelier et al. Downloaded by guest on October 1, 2021